An Analysis of Lesion Size and Location in Newly Diagnosed Cytomegalovirus Retinitis

Ramin Sarrazadeh, MD, PhD,1 David V. Weinberg, MD,1 Cheng-Fang Huang, MS2
Objective: To investigate the size and distribution of lesions in newly diagnosed cytomegalovirus retinitis (CMVR). Design: Retrospective, observational case series. Participants: Fundus photographs of 252 newly diagnosed CMVR lesions in 173 eyes of 130 patients (123 male, 7 female). Methods: Thirty-ve millimeter (60) color transparencies were digitized. A montage of the retina was assembled for each involved eye and was superimposed on a specially designed map of the postequatorial retina. Cytomegalovius retinitis lesions were delineated, and the size and location of each lesion was measured. The size of newly diagnosed CMVR lesions was computed in terms of percent postequatorial retinal surface area (PERSA), and the location of lesions was plotted on a polar coordinate system. Main Outcome Measures: Size and location of patches of newly diagnosed CMVR. Results: The median lesion size was 3% PERSA. Peripheral CMVR lesions were larger than posterior ones (P 0.001). The mean number of lesions was 1.6 per eye. The total area of CMVR involvement ranged from 1% to 76% PERSA, with a median of 5% PERSA. There was no difference between left and right eyes in the distribution of lesion centers (P 0.27). The concentric distribution of lesion centers appeared to be homogeneous, except for fewer centers in the most peripheral 14 (P 0.001), and a greater than expected number of lesion centers in the macula (central 11.6, P 0.001). Eyes of patients with unilateral retinitis had 1.3 lesions per involved eye compared with eyes of patients with bilateral retinitis, which had 1.6 lesions per eye (P 0.003). Conclusions: Most newly diagnosed CMVR lesions were small. Peripheral lesions were larger than more posterior lesions. Variations from a homogeneous distribution of lesions were noted only at the extreme peripheral and central locations and are probably explained by ascertainment bias. The macula was not spared from new CMVR lesions in this patient population. Ophthalmology 2002;109:119 125 2002 by the American Academy of Ophthalmology. Cytomegalovirus retinitis (CMVR), the most common opportunistic ocular infection seen in patients with acquired immunodeciency syndrome (AIDS), is a major cause of visual morbidity in patients with AIDS.13 Loss of vision attributable to CMVR is inuenced by the size and location of the lesions. Cytomegalovirus can damage vision in several ways; it can infect and destroy the macula or optic nerve directly,4 6 or it can infect the retinal periphery, resulting in loss of visual eld.712 Large, peripherally located CMVR lesions increase the risk of retinal detachment, another cause of visual morbidity in eyes with CMVR.10 12 Cytomegalovirus retinitis lesion size and location have been investigated in a number of ways in previous studies. The location of CMVR lesions has been studied3,5,10,13 using a standard zone grading system derived by Holland et al,14 which divides the retina into three concentric zones. Zone I is dened as including all structures within 1500 m of the optic nerve head, 3000 m of the fovea, or both. Zone II extends from zone I to a circle drawn through the ampullae of the vortex veins, and zone III demarcates the retina anterior to zone II. This scheme was adopted by the Studies of Ocular Complications of AIDS (SOCA) research group. Lesion size has been estimated using the linear dimensions of the optic nerve13 or by clinically estimating the retinal involvement in terms of percent retinal surface area.10 Some studies have concluded that CMVR obeys a macula sparing pattern, with most lesions being located in the periphery.1,6 In contrast, other studies have found signicant CMVR involvement of the posterior retina.3,5,9,15 In the present study, we sought to analyze quantitatively the size and location of lesions in newly diagnosed CMVR to determine the relative extent of involvement among different parts of the retina. Because photographs are a at representation of a curved surface, image distortion is inISSN 0161-6420/02/$see front matter PII S0161-6420(01)00878-8

Originally received: November 6, 2000. Accepted: April 24, 2001. Manuscript no. 200764. 1 Department of Ophthalmology, Northwestern University Medical School, Chicago, Illinois. 2 Department of Preventative Medicine, Northwestern University Medical School, Chicago, Illinois. Presented in part as a poster at the Association for Research in Vision and Ophthalmology annual meeting, Fort Lauderdale, Florida, May 2000. Supported by an unrestricted grant from Research to Prevent Blindness, Inc., New York, New York. The authors have no conicting commercial interests. Correspondence to David V. Weinberg, MD, Department of Ophthalmology, Northwestern University Medical School, 645 North Michigan Avenue, Suite 440, Chicago, IL 60611. E-mail: d-weinberg2@northwestern. edu.
2002 by the American Academy of Ophthalmology Published by Elsevier Science Inc.

119

Ophthalmology Volume 109, Number 1, January 2002

tino, CA). The polar map was constructed and the CMVR lesion data graphically plotted using the ClarisDraw graphics program 1.0 version 2 (Claris Corp., Santa Clara, CA) and the Canvas graphics software program version 5.0 (Deneba Software, Miami, FL).

Data Analysis
Each image was scaled so that the horizontal extent of each photograph spanned 60 on the map. The anterior borders of the ampullae of the vortex veins were found to extend just posterior to the equator of the map, conrming that the images had been scaled correctly. For each eye, a photographic montage of the fundus was created and superimposed on the at map. Anatomic landmarks such as retinal blood vessels were used to align the images in the montage. The images were enlarged on a high-resolution 17-inch color monitor to mark the borders of CMVR. Both the active and inactive regions of CMVR were included in each lesion. Small, satellite lesions adjacent to dense, white areas of active retinitis were considered part of the lesion and were included within the border. The original transparencies were used to help identify lesion borders when they could not be identied denitively from the digitized images. Lesion centers were determined by the point of intersection of lines drawn across the longest dimension, and the line drawn perpendicular to the longest dimension of the lesion. The distance from the fovea was determined in degrees for the center, anterior, and posterior border of all lesions, and polar coordinates for all lesion centers were computed. Analysis of the complete extent of each lesion was such that if part of any lesion encroached into a region, it was said to involve that region. Lesion size was determined by further subdividing the polar map into regions of 0.25% PERSA. Lesion size was then graphically estimated and rounded to the nearest whole number percent PERSA for each lesion. For all lesions, the color fundus photos were compared with detailed retina drawings on clinic notes. Six CMVR lesions were located anterior to the equator in their entirety. These lesions were excluded from this study. The complete CMVR lesion border had been photographed for 204 lesions included in this study; 48 additional lesions were photographed to the equator; however, clinic records indicated that the lesions actually extended anterior to the equator. Lesion centers and surface areas were computed for these lesions by assuming that they ended at the equator. The data set was analyzed by rst including and then excluding these additional 48 lesions. The complete postequatorial distribution of all 252 lesions was then analyzed. For the overall analyses of lesion size and location, each lesion was treated as an independent event. Two lesions from one eye were counted the same as lesions from two different eyes or two different patients. Data analysis was performed using statistical software (SAS Institute Inc., Cary, NC). Unless otherwise specied, categorical data were analyzed by chi-square test and continuous data by a two-tailed t test or the Wilcoxon rank-sum test. Statistical significance was considered P 0.05.

Figure 1. The postequatorial retina depicted as the inner surface of a hemisphere, divided into 49 regions of equal surface area.

herent and confounds quantitative analysis. We designed a mapping system that compensates for some of the photographic distortion and that permits analysis of the concentric as well as the circumferential distribution of CMVR lesions and their size.

Materials and Methods

Construction of the Retinal Map
The postequatorial retina was represented as a hemisphere. The retinal surface was divided into 49 regions of equal area, each region containing 2.04% of the total postequatorial retinal surface area (PERSA; Fig 1). The central 11.6 surrounding the fovea is dened as the macula region. The macula is surrounded by 4 concentric zones, and each zone is subdivided into 12 regions. The zone immediately surrounding the macula is referred to as the arcade zone. The most peripheral zone is referred to as the postequatorial zone. This hemisphere was then transposed into a at map to account for the distortion inherent in a at photographic representation of a curved surface. For further details on the construction of the retinal map, please see the Appendix.

Data Selection
Data selection was performed by a chart review of all patients with newly diagnosed CMVR seen by one of the investigators (DVW) at the Department of Ophthalmology at Northwestern University Medical School from 1990 through 1997. The medical records and the Kodachrome color transparencies (Eastman Kodak Co., Rochester, NY) obtained at the time of diagnosis were then studied.

Data Acquisition
Kodachrome color transparencies taken at the time of diagnosis of CMVR using a Canon model CF-60UV 60 fundus camera (Canon U.S.A., Lake Success, NY) were digitized with a Polaroid SprintScan 35 (Polaroid Corp., Cambridge, MA) slide scanner using Adobe Photoshop version 4.0 (Adobe Systems Inc., Mountain View, CA). The images were then transferred to an Apple Computer PowerMacintosh G3/220 (Apple Computer Corp., Cuper-

Results
Two hundred fty-two lesions in 173 eyes of 130 patients (123 male, 7 female) with newly diagnosed CMVR were analyzed. There were 122 right eye lesions and 130 left eye lesions. Cytomegalovirus retinitis lesions were present in both eyes in 43 of 130 patients (33%).

120

Sarrazadeh et al Lesion Size and Location in CMV Retinitis

Figure 2. An azimuth-equidistant projection of the retina model depicted in Figure 1. Fundus montages were superimposed on this map for determination of the location and size of cytomegalovirus retinitis lesions. Polar plots illustrating the distribution of lesion centers for all eyes, as well as the distribution of the total surface area of all lesions for all eyes, are shown. A, A right eye adjusted polar plot of all lesion centers for all eyes. The coordinates of the lesion centers for left eyes were adjusted to match nasal and temporal orientation on a map of the right eye. The total number of lesion centers in each region and quadrant are shown to the right. B, A right eye-adjusted polar plot of the distribution of the total surface area of all lesions. If a lesion encroached into any part of a region, it was said to involve that region. The map for left eyes was adjusted to match nasal and temporal lesions onto the map for right eyes. The distribution by zone is shown to the left, whereas the involvement by region and quadrant is illustrated to the right.

There was no detectable difference in the distribution of lesions among regions between left and right eyes (chi-square test, P 0.27). The distribution of lesion centers for all eyes and the distribution of the complete surface area of all lesions for all eyes

are shown in Figure 2. Analysis of the quadrantic location of lesions showed no detectable difference from a uniform distribution for lesion centers (P 0.71) or for any part of a lesion (P 0.76). The distribution of lesion centers over the 49 regions was

121

Ophthalmology Volume 109, Number 1, January 2002

Figure 3. A frequency histogram depicting the size distribution of individual cytomegalovirus retinitis lesions (CMVR). The mean and median size was 4.8% and 3% postequatorial retinal surface area (PERSA), respectively. Fewer than one tenth of the lesions were 10% or more PERSA. Figure 4. A scatterplot of lesion size versus distance from lesion center to the fovea. The regression line demonstrates a positive correlation (Pearson correlation, r 0.21, P 0.0006).

not uniform (chi-square test for equal proportions, P 0.001). The most obvious variation from a uniform distribution was the presence of relatively few lesions in the postequatorial zone. Analysis of the more posterior 37 regions (macula, arcade zone, midperipheral zone, and peripheral zone) suggested a trend away from a uniform distribution, which does not reach the traditional level of statistical signicance (P 0.09). Figure 2A summarizes the distribution of the number of lesion centers among all the regions. The sole outlier (more than two standard deviations from the mean) is for the central macula region. This region accounts for 2% PERSA, but contains 6% of the lesion centers. The greater number of lesions in the macula region was statistically signicant (P 0.001, goodness-of-t chi-square test). If the postequatorial zone and the macular regions are excluded, the distribution of lesion centers among the remaining 36 regions showed no detectable difference from what would be predicted by a homogeneous distribution (P 0.33). Figure 3 displays the distribution of lesion size. The distribution is skewed strongly toward small lesions. The median size was 3% PERSA. Fewer than one tenth of the lesions were 10% or more PERSA. Involved eyes contained up to six lesions, but most had just one lesion. The mean number of lesions was 1.6 per eye. The total area of CMVR involvement (combined PERSA for all lesions in an eye) ranged from 1% to 76% PERSA, with a median of 5% PERSA. There was a positive correlation between distance from the fovea and lesion size (Pearson correlation, r 0.21, P 0.0006; Fig 4). Lesions with centers within the median distance from the fovea were 3.8% PERSA in mean size, whereas lesions with centers outside the median were 5.9% PERSA in mean size (P 0.002). Eyes of patients with unilateral retinitis had 1.3 lesions per involved eye compared with eyes of patients with bilateral retinitis, which had 1.6 lesions per eye (P 0.003). The mean posterior lesion border was 24 from the fovea in unilaterally involved eyes compared with 28 in bilaterally involved eyes, a difference of borderline statistical signicance (P 0.05). Comparing eyes of patients with unilateral versus bilateral retinitis, there was no difference in the distance of lesion centers from the fovea (P 0.38), individual lesion area (P 0.11), or total area of retinitis per involved eye (P 0.98).

Discussion
The present study provides a quantitative evaluation of newly diagnosed CMVR lesions by using a mapping system that simplied the analysis of lesion location and size and allowed statistical analysis. The results suggest that newly diagnosed CMVR is uniformly distributed over most of the retinal surface. Variation from a uniform distribution was found in the postequatorial periphery, where there was a relative scarcity of lesion centers, and in the macula, which contained far more lesion centers than would be predicted by its relative surface area. Before considering the potential signicance of these results, we rst discuss the advantages as well as the limitations of the methods used.

Suitability of the Polar Map for Studying Cytomegalovirus Retinitis Lesions

There are a number of difculties in quantitatively analyzing retinal lesions using fundus photographs. First, fundus imaging systems have variable magnication characteristics that are not only a function of the camera used, but also the refractive power of the eye photographed.16,17 Second, sizes and distances may be distorted, particularly for objects approaching and anterior to the equator. Third, not all CMVR lesions can be photographed completely, especially those extending anterior to the equator. Prior studies have outlined clever ways of measuring the optic nerve dimensions or the size of small posterior pole lesions using multiple-beam interference fringes18 or a projected ruler.19 These techniques, unfortunately, are not suitable for studying geometrically complex CMVR lesions that may extend to the midperiphery or beyond. In the present study, we attempted to devise a polar map based on the curved surface area of the retina. This was then transposed to a at representation simulating how the retina

122

Sarrazadeh et al Lesion Size and Location in CMV Retinitis

would be depicted in photomontages constructed from fundus photographs. Similar methods have been described previously.20,21 Ideally, one would construct a different map for each eye taking into account its refractive power. Unfortunately, this is not feasible. The model we designed will slightly overestimate the retinal surface area in some patients and will undoubtedly underestimate it in others, depending on the refractive characteristics of the eyes. All patients included in this study had been photographed on the same camera, thereby providing some uniformity in the quality as well as magnication characteristics of the images. The construction of our map assumes that the postequatorial retina is a spherical surface with the macula at its apex and that the photomontages prepared from clinical photographs represent an azimuth equidistant representation of this spherical surface.21 We divided the postequatorial retina into 49 equally sized regions and transposed the boundaries of these regions as they would be represented on such a at map. By measuring lesion area as a proportion of the whole (% PERSA), we hoped to avoid the technical difculties inherent in trying to measure the absolute size of objects from fundus photographs. Because lesions anterior to the equator are difcult to photograph and are subject to the greatest distortion, we limited our study to the distribution of lesions in the postequatorial retina. This method of analysis is reliable for studying the complete postequatorial distribution of CMVR lesions but may have some limitations with respect to determining the location of lesion centers for the 48 incompletely photographed lesions. For 204 completely photographed lesions, the lesion center could be determined with reasonable accuracy. For the other 48 lesions, it was known that the true peripheral border of the lesion was anterior to the equator, however, this border was not reliably documented. Eliminating these lesions from the data pool would have biased the data greatly toward more posterior lesions, because most of these 48 lesions were anteriorly located. Instead, these lesions were measured by assuming that each lesion ended at the equator, and these data were included in all reported analyses. Although the postequatorial retinal surface area calculations for these lesions are valid, the lesion centers determined in this manner are more posterior than they are in reality. This is a source of bias toward more posterior localization of lesions, but this bias clearly was less than if these lesions had been entirely excluded from the analysis. Furthermore, analysis of the complete surface area of all lesions conrmed the results obtained with lesion center analysis, demonstrating fewer lesions in the postequatorial zone. It is possible that some lesions that we counted as one may have resulted from two or more lesions that had coalesced. This would have resulted in undercounting of lesions and may have biased our results. Because the median lesion size was small, we believe it is unlikely that we undercounted a signicant number of lesions in this way.

Quantitative and Topographic Features of Newly Diagnosed Cytomegalovirus Retinitis

In our study, the far peripheral retina (14 posterior to the equator) showed a relative scarcity of lesion centers. The distribution of lesions in the remainder of the postequatorial retina appeared homogeneous, with the exception of the 2% of retina centered in the macula, which contained more lesion centers than its contribution to the retinal surface area would predict. Some descriptions of CMVR have suggested that the infection has a predilection to begin around the major vascular arcades, as may be expected based on presumed hematogenous seeding of the virus. However, quantitative analysis did not demonstrate any greater number of lesion centers in the arcade zone compared with the surrounding midperipheral and peripheral zones. Previous reports of the distribution of CMVR lesions have been predominantly descriptive or semiquantitative and have not taken into consideration the relative area of the different parts of the retina. Furthermore, these studies have not accounted for the image distortion inherent to the various methods of data acquisition that were used. Some authors have reported that posterior retinal involvement is a prominent feature of CMVR,35 whereas others have emphasized the involvement of the periphery.1,6,15 The SOCA research group reported that in 337 eyes with newly diagnosed CMVR, 50% showed zone I involvement, 96% showed zone II involvement, and 84% showed zone III involvement. Involvement of the optic nerve (9%) or fovea (4%) occurred infrequently.22 The SOCA group uses a standard fundus photography protocol and a centralized reading center. The mapping system used by SOCA divides the retina into three zones based on fundus landmarks, which makes it convenient for clinical classication of lesion location but is much less suitable for quantitative analysis. The extent of retinal involvement is measured by planimetry from the transparencies. It is not clear if adjustments are made for image distortion of peripheral retinal lesions. Furthermore, SOCA has not published data localizing lesions by a unique coordinate, such as lesion center, so only limited comparisons of our results with those reported by SOCA are possible. The total area of retinal involvement in percent PERSA was somewhat less in our patients than that reported for a similar cohort by the SOCA research group, which found a mean of 13.5% PERSA for unilaterally involved eyes.22 Bilaterally involved eyes were analyzed as the better or worse eyes. The mean percent PERSA was found to be 7.9% in the better eyes and 20.4% in the worse eyes in the SOCA study. It appears that the bilateral eyes overall did not differ from unilateral eyes, however, no analysis of such a comparison was reported.22 We found the mean total area for all involved eyes to be 10.3% PERSA, with no difference between eyes for unilaterally versus bilaterally involved patients. In this paper, we present mean area only as a way of comparing our data with SOCA. We believe that median is a more representative measure of central tendency in this asymmetric distribution. We found an overall distribution with more lesions located in the posterior than the peripheral zones. The more

123

Ophthalmology Volume 109, Number 1, January 2002

posterior relative distribution of lesions in this analysis was driven chiey by the rare occurrence of lesions in the most peripheral postequatorial zone. When this zone was excluded from the analysis, there was no signicant posterior predisposition of lesions, with the exception of the macula region. We also found that peripheral lesions were, on the average, larger than more posterior lesions. There are several possible reasons for these ndings. It is possible that CMVR is less likely to seed in the more peripheral retina. However, it is also likely that there is an ascertainment bias resulting in the exclusion of more peripheral lesions. Peripheral lesions are less likely to result in symptoms for the patient and therefore are less likely to lead to a clinical diagnosis of CMVR. As a result, peripheral lesions may be underrepresented in any clinical collection of CMVR patients. Also, far peripheral lesions are more difcult to document with photographs, so some peripheral lesions in our patients may have escaped analysis. Posterior lesions are also more likely to result in symptoms at a relatively small size, whereas peripheral lesions are likely to be larger before they cause symptoms or are discovered on screening examination. Quantitative analysis showed the macula region to be involved more frequently than any other region and triple that which would be predicted based on its relative proportion of the retinal surface area. An ascertainment bias exists because macular lesions are likely to be symptomatic early in their course, however, early CMVR clearly does not occur in a macula-sparing distribution in our patient sample. Eyes of bilaterally infected patients contained more lesions per eye than those of unilaterally infected patients. Bilateral involvement may occur in a setting of more extreme immune suppression, of more severe CMV viremia, or both, with multiple foci of retinal seeding. One of the potential uses of our data are in the design of screening protocols for CMVR. Because visualization of any portion of a CMV lesion would be a positive result, it is the posterior border that would be of greatest relevance in determining the sensitivity of a screening technique. The median posterior extent of lesions was 23.5; thus, to detect 50% of lesions, at least 23.5 from the fovea would need to be visualized. To detect 75% of lesions, 38 would need to be visualized. To increase the sensitivity to 95%, visualization of 59 in all directions from the fovea would be required. It should be recognized that these data probably overestimate the sensitivity of screening. Our data include symptomatic as well as asymptomatic patients. Screening is of the greatest value in detecting retinitis in asymptomatic patients, whose lesions are likely to be more peripheral than in our cohort. In summary, we found CMVR to be evenly distributed through the postequatorial retina except in the extreme periphery, where fewer lesions were centered, and in the most central macular region, where an excess number of lesion centers were found. Peripheral lesions were larger than more posterior lesions. Ascertainment bias is a likely factor in these variations from a uniform distribution. These data add to our basic knowledge concerning the clinical behavior of CMV in the retina. The techniques developed for this study may be of value in studying the topographic features of other retinal landmarks and diseases.

References
1. Egbert PR, Pollard RB, Gallagher JG, Merigan TC. Cytomegalovirus retinitis in immunosuppressed hosts. II. Ocular manifestations. Ann Intern Med 1980;93:664 70. 2. Khadem M, Kalish SB, Goldsmith J, et al. Ophthalmologic ndings in acquired immune deciency syndrome (AIDS). Arch Ophthalmol 1984;102:201 6. 3. Jabs DA, Enger C, Bartlett JG. Cytomegalovirus retinitis and acquired immunodeciency syndrome. Arch Ophthalmol 1989;107:75 80. 4. Gross JG, Sadun AS, Wiley CA, Freeman WR. Severe visual loss related to isolated peripapillary retinal and optic nerve head cytomegalovirus infection. Am J Ophthalmol 1989;108: 691 8. 5. Gross JG, Bozzette SA, Mathews WC, et al. Longitudinal study of cytomegalovirus retinitis in acquired immune deciency syndrome. Ophthalmology 1990;97:681 6. 6. Glasgow BJ, Weisberger AK. A quantitative and cartographic study of retinal microvasculopathy in acquired immunodeciency syndrome. Am J Ophthalmol 1994;118:46 56. 7. Broughton WL, Cupples HP, Parver LM. Bilateral retinal detachment following cytomegalovirus retinitis [case report]. Arch Ophthalmol 1978;96:618 9. 8. Meredith TA, Aaberg TM, Reeser FH. Rhegmatogenous retinal detachment complicating cytomegalovirus retinitis [case report]. Am J Ophthalmol 1979;87:793 6. 9. Freeman WR, Henderly DE, Wan WL, et al. Prevalence, pathophysiology, and treatment of rhegmatogenous retinal detachment in treated cytomegalovirus retinitis. Am J Ophthalmol 1987;103:52736. 10. Jabs DA, Enger C, Haller J, de Bustros S. Retinal detachments in patients with cytomegalovirus retinitis. Arch Ophthalmol 1991;109:794 9. 11. Freeman WR, Friedberg DN, Berry C, et al. Risk factors for development of rhegmatogenous retinal detachment in patients with cytomegalovirus retinitis. Am J Ophthalmol 1993; 116:71320. 12. Rhegmatogenous retinal detachment in patients with cytomegalovirus retinitis: the Foscarnet-Ganciclovir Cytomegalovirus Retinitis Trial. The Studies of Ocular Complications of AIDS Research Group in collaboration with the AIDS (SOCA) Clinical Trials Group. Am J Ophthalmol 1997;124:6170. 13. Holland GN, Shuler JD. Progression rates of cytomegalovirus retinopathy in ganciclovir-treated and untreated patients. Arch Ophthalmol 1992;110:1435 42. 14. Holland GN, Buhles WC, Mastre B, et al. A controlled retrospective study of ganciclovir treatment for cytomegalovirus retinopathy. Use of a standardized system for the assessment of disease outcome. UCLA CMV Retinopathy Study Group. Arch Ophthalmol 1989;107:1759 66. 15. Henderly DE, Freeman WR, Causey DM, Rao NA. Cytomegalovirus retinitis and response to therapy with ganciclovir. Ophthalmology 1987;94:42534. 16. Rudnicka AR, Burk ROW, Edgar DF, Fitzke FW. Magnication characteristics of fundus imaging systems. Ophthalmology 1998;105:2186 92. 17. Garway-Heath DF, Rudnicka AR, Lowe T, et al. Measurement of optic disc size: equivalence of methods to correct for ocular magnication. Br J Ophthalmol 1998;82:6439.

124

Sarrazadeh et al Lesion Size and Location in CMV Retinitis

18. Baumbach P, Rassow B, Wesemann W. Absolute ocular fundus dimensions measured by multiple-beam interference fringes. Invest Ophthalmol Vis Sci 1989;30:2314 9. 19. Arzabe CW, Jalkh AE, Fariza E, et al. A simple device to standardize measurements of retinal structures in fundus photographs and retinal angiograms. Am J Ophthalmol 1990;109: 107 8. 20. Mosier MA. Retinal cartography. Can J Ophthalmol 1982;17: 219 22. 21. Borodkin MJ, Thompson JT. Retinal cartography: an analysis of two-dimensional and three-dimensional mapping of the retina. Retina 1992;12:273 80. 22. Foscarnet-Ganciclovir Cytomegalovirus Retinitis Trial: 5. Clinical features of cytomegalovirus retinitis at diagnosis. Studies of Ocular Complications of AIDS Research Group in collaboration with the AIDS Clinical Trials Group. Am J Ophthalmol 1997;124:14157.

P A c / A h 1 cos. Solving for , cos 1 1 P ). In our map, we dene a central cap to represent 1/49 the total area of the hemisphere, such that: 1 cos 1 1 1/49 11.6 . The remaining polar caps are dened with P of 13/49, 25/49, 37/49, and 49/49, resulting in s of 42.7, 60.7, 75.8, and 90, respectively. Each band surrounding the central cap is divided into 12 equally sized segments, resulting in a total of 49 equally sized regions, each representing 1/49 or approximately 2% PERSA. Each of the 49 segments of retina is referred to as a regions, and each of the 4 bands of retina surrounding the central region is referred to as a zone composed of 12 regions. The central region is referred to as the macula, the innermost zone surrounding the macula is called the arcade zone, the next the midperipheral zone, then the peripheral zone, and the outermost as the postequatorial zone. Position on the map was dened using a system analogous to latitude and longitude on geographic maps. The distance from the center of the fovea was measured in degrees (0 at the center of the fovea to 90 at the equator). Circumferential location was identied by designating the 3-oclock meridian as 0 and using circular coordinates in a counterclockwise orientation (i.e., 12-oclock is 90, 9-oclock is 180, 6-oclock is 270, etc.). To account for the distortion inherent in a photographic montage of a curved surface, a at, equatorial, equidistant, projection of the polar map was constructed.20,21 In this projection of a hemisphere, angular distances from the pole (fovea) are directly proportional to linear distance. The circumferential orientation (i.e., clock-hour location) is preserved, however, the apparent circumference becomes increasingly exaggerated or stretched toward the equator.

Appendix
Construction of the Retinal Map
A hemispheric polar map centered on the fovea was constructed that divided the postequatorial retina into 49 regions of equal surface area, each region representing approximately 2% of the total postequatorial retinal surface area (PERSA). The map was composed of a central region, surrounded by four concentric zones each divided into 12 regions (Fig 1). The postequatorial retina was represented by the inner surface of a hemisphere with the fovea at its apex. The surface area of a hemisphere (Ah) is: A h 2r 2. The area of a polar cap (Ac) of the hemisphere is A c 2r 2(1 cos), where represents the angle between radii drawn to the apex and to the edge of the cap. Let P represent the ratio of the area of the spherical cap relative to the area of the entire hemisphere.

125